机载激光雷达数据中道路中线的多尺度提取方法

Multi-scale method for extracting road centerlines from LIDAR datasets

机载激光雷达(LIDAR)技术的出现为道路特征的获取提供了新的途径。在分析现有道路提取现状的基础上,针对激光雷达数据的特点以及单一尺度下道路中线提取方法的不足,提出一种基于多尺度追踪的道路中线提取方法。该方法首先采用逐步约束的方法进行道路激光点的提取,包括高程约束、强度约束以及区域点密度和区域面积的约束等;然后基于道路点云生成的不同尺度距离影像的形态学细化结果,采用多尺度追踪的方法实现道路中线的提取,其中多尺度追踪方法由大尺度道路中线的迭代追踪以及小尺度道路中线的启发式追踪两部分组成。最后采用实地数据进行验证,结果表明:该方法能有效地从LIDAR点云中提取道路中线信息,并具有较好的精度。

The LIDAR technology provides a new approach for acquiring road information.Having analyzed characteristics of LIDAR datasets as well as deficiencies of current road extracting methods, a multi-scale tracing method was proposed to extract the road centerlines. Firstly, candidate road points were segmented from raw LIDAR points cloud by a series of constraint procedures including height- constraint, intensity-constraint, local point density and region area constraint, and so on. Secondly, based on morphological thinning results of multi-scale road range images generated from road points segmented above, road centerlines were extracted by a multi-scale centerline tracing method, which consisted of two steps：an iterative tracing procedure on the large scale range image and a heuristics tracing procedure on the small one. Finally, the proposed method was validated by one LIDAR dataset with complex road networks. The result shows that the proposed method can extract road centerlines from LIDAR points cloud effectively and precisely.